Some theoretical aspects of support vector machines and related kernel-based learning methods

Ingo Steinwart
Los Alamos National Laboratory
Machine Learning and Pattern Recognition Team

Support vector machines (SVMs) and related algorithms are used for a variety of learning scenarios including
classification and regression. This talk mainly focuses on some recent progress in the theoretical analysis of these
algorithms for both classification and regression problems. In particular, we will discuss in which sense SVMs can learn and which additional
features they typically have in the learning process. If time permits we finally discuss the role of the used loss function in some more detail.


Presentation (PDF File)
Video of Talk (RealPlayer File)

Back to Graduate Summer School: Intelligent Extraction of Information from Graphs and High Dimensional Data